My Gemini Live API based assistant for an experimental CX journey can now read barcodes when conversing with the customer.
1. Customer establishes video WebSocket stream
2. Server stores the last X seconds video frames
3. Model calls tool which sends frames to barcode detector
00:00

